Comparison of conventional 2D ultrasound to magnetic resonance imaging for prenatal estimation of birthweight in twin pregnancy

Background During prenatal follow-up of twin pregnancies, accurate identification of birthweight and birthweight discordance is important to identify the high-risk group and plan perinatal care. Unfortunately, prenatal evaluation of birthweight discordance by 2-dimensional ultrasound has been far from optimal. Objective The objective of the study was to prospectively compare estimates of fetal weight based on 2-dimensional ultrasound (ultrasound–estimated fetal weight) and magnetic resonance imaging (magnetic resonance–estimated fetal weight) with actual birthweight in women carrying twin pregnancies. Study Design Written informed consent was obtained for this ethics committee–approved study. Between September 2011 and December 2015 and within 48 hours before delivery, ultrasound–estimated fetal weight and magnetic resonance–estimated fetal weight were conducted in 66 fetuses deriving from twin pregnancies at 34.3–39.0 weeks; gestation. Magnetic resonance–estimated fetal weight derived from manual measurement of fetal body volume. Comparison of magnetic resonance–estimated fetal weight and ultrasound–estimated fetal weight measurements vs birthweight was performed by calculating parameters as described by Bland and Altman. Receiver-operating characteristic curves were constructed for the prediction of small-for-gestational-age neonates using magnetic resonance–estimated fetal weight and ultrasound–estimated fetal weight. For twins 1 and 2 separately, the relative error or percentage error was calculated as follows: (birthweight – ultrasound–estimated fetal weight (or magnetic resonance–estimated fetal weight)/birthweight) × 100 (percentage). Furthermore, ultrasound–estimated fetal weight, magnetic resonance–estimated fetal weight, and birthweight discordance were calculated as 100 × (larger estimated fetal weight–smaller estimated fetal weight)/larger estimated fetal weight. The ultrasound–estimated fetal weight discordance and the birthweight discordance were correlated using linear regression analysis and Pearson's correlation coefficient. The same was done between the magnetic resonance–estimated fetal weight and birthweight discordance. To compare data, the χ 2, McNemar test, Student t test, and Wilcoxon signed rank test were used as appropriate. We used the Fisher r-to-z transformation to compare correlation coefficients. Results The bias and the 95% limits of agreement of ultrasound–estimated fetal weight are 2.99 (–19.17% to 25.15%) and magnetic resonance–estimated fetal weight 0.63 (–9.41% to 10.67%). Limits of agreement were better between magnetic resonance–estimated fetal weight and actual birthweight as compared with the ultrasound–estimated fetal weight. Of the 66 newborns, 27 (40.9%) were of weight of the 10th centile or less and 21 (31.8%) of the fifth centile or less. The area under the receiver-operating characteristic curve for prediction of birthweight the 10th centile or less by prenatal ultrasound was 0.895 (P <.001; SE, 0.049), and by magnetic resonance imaging it was 0.946 (P <.001; SE, 0.024). Pairwise comparison of receiver-operating characteristic curves showed a significant difference between the areas under the receiver-operating characteristic curves (difference, 0.087, P =.049; SE, 0.044). The relative error for ultrasound–estimated fetal weight was 6.8% and by magnetic resonance–estimated fetal weight, 3.2% (P <.001). When using ultrasound–estimated fetal weight, 37.9% of fetuses (25 of 66) were estimated outside the range of ±10% of the actual birthweight, whereas this dropped to 6.1% (4 of 66) with magnetic resonance–estimated fetal weight (P <.001). The ultrasound–estimated fetal weight discordance and the birthweight discordance correlated significantly following the linear equation: ultrasound–estimated fetal weight discordance = 0.03 + 0.91 × birthweight (r = 0.75; P <.001); however, the correlation was better with magnetic resonance imaging: magnetic resonance–estimated fetal weight discordance = 0.02 + 0.81 × birthweight (r = 0.87; P <.001). Conclusion In twin pregnancies, magnetic resonance–estimated fetal weight performed immediately prior to delivery is more accurate and predicts small-for-gestational-age neonates significantly better than ultrasound–estimated fetal weight. Prediction of birthweight discordance is better with magnetic resonance imaging as compared with ultrasound.